Nonequilibrium Kinetic Model That Describes the Reversible

Feb 13, 2007 - and desorption cycles carried out at 400 °C. A long cycle time adsorption (700 ..... adsorption-desorption run over the entire 1400-mi...
1 downloads 0 Views 162KB Size
Download PDF
Ind. Eng. Chem. Res. 2007, 46, 1737-1744

1737

MATERIALS AND INTERFACES Nonequilibrium Kinetic Model That Describes the Reversible Adsorption and Desorption Behavior of CO2 in a K-Promoted Hydrotalcite-like Compound Armin D. Ebner, Steven P. Reynolds, and James A. Ritter* Department of Chemical Engineering, UniVersity of South Carolina, Columbia, South Carolina 29208

A nonequilibrium kinetic model was developed to describe the reversible adsorption and desorption behavior of CO2 in a K-promoted hydrotalcite-like compound (HTlc). The model consisted of three reversible reactions. Two of the reactions were of the Langmuir-Hinshelwood type with slow and intermediate kinetics, and one was a mass-transfer-limited chemisorption process with very fast kinetics. To calibrate and test this model, a K-promoted HTlc was synthesized and studied to determine its dynamic behavior during CO2 adsorption and desorption cycles carried out at 400 °C. A long cycle time adsorption (700 min) and desorption (700 min) experiment was carried out with a sample activated at 400 °C for 12 h in helium. With this experiment approaching equilibrium at the end of each step, it proved that the adsorption and desorption behavior of CO2 in K-promoted HTlc was completely reversible. Then, the effect of the half-cycle time (15, 30, 45, 60, and 75 min) was studied with samples activated for 12 h in helium at 400 °C and cycled four times each, and the effect of the activation time (8, 12, 16, and 20 h) was studied with samples cycled twice with a 45-min half-cycle time. The former set of experiments proved that periodic behavior was achieved very quickly with cycling even when far removed from an equilibrium state; the latter set proved that the CO2 working capacity was independent of the activation time. The model was fitted successfully to the long cycle time experiment. It then predicted successfully the dynamic and cyclic behavior of both the much shorter cycle time and different activation time experiments. This kinetic model accurately simulated the reversible adsorption and desorption behavior of the very fast, intermediate, and slow kinetic processes; the approach to periodic behavior during cycling; and the independence between the CO2 working capacity and activation time. It also proved that the adsorption and desorption behavior was due to a combination of completely reversible adsorption, diffusion, and reaction phenomena. Introduction Hydrotalcite-like compounds (HTlcs) that exhibit a reversible capacity for CO2 at elevated temperatures are being explored for the removal of CO2 from equilibrium limited reactions1-5 and for the capture and concentration of CO2 from flue gas.6-12 Nevertheless, a paucity of literature is available on the adsorption properties of CO2 on HTlcs.1-16 A mechanism that clearly describes the reversible CO2 uptake and release processes is also sorely lacking. Much of the literature that describes the reversible adsorption of CO2 on HTlcs treats it as a hightemperature, diffusion-limited, equilibrium driven process that is akin to physical adsorption.4-12 However, on the basis of discrepancies in the values of the mass-transfer coefficients of CO2 in HTlcs reported in the litereature,2,4-10 the actual mechanism appears to be much more complicated than this simple depiction. For example, Ding and Alpay4 reported mass-transfer coefficients for CO2 in a K-promoted HTlc of 0.0058 s-1 for adsorption and 0.0006 s-1 for desorption. Those values both indicated the dominance of a slow-diffusion or reaction-limited process. In contrast, Hufton et al.2 showed very steep breakthrough curves and an elution curve predicted by equilibrium theory for CO2 in a K-promoted HTlc. Very fast mass transfer * To whom correspondence should be addressed. Tel.: (803) 7773590. Fax: (803) 777-8365. E-mail: [email protected].

or reaction kinetics were indicated by those behaviors. Finally, Soares et al.9 reported a mass-transfer coefficient for CO2 in HTlcs as high as 0.0153 s-1. That value indicated a mass-transfer or reaction-limited process with CO2 uptake and release rates lying somewhere in-between those reported in the other studies. To further complicate matters, recent results10 suggested that the reversible adsorption of CO2 on HTlcs at elevated temperatures was a kinetically driven, nonequilibrium process that acquires the character of an equilibrium process only after extremely long times. Ebner et al.16 showed that an initially fast adsorption or desorption phenomenon was followed by a state of extremely slow uptake or release of CO2 that requires hours perhaps days to reach equilibrium. The CO2 loadings at such an equilibrium state were also considerably different than those attained during the initial stages of adsorption or desorption. This kind of unusual behavior of CO2 on HTlcs at elevated temperatures was consistent with the Langmuir-Hinshelwood type of kinetic model mentioned recently by Moreira et al.10 Therefore, the objective of this article is to report on the development of a nonequilibrium kinetic model that describes the reversible adsorption and desorption behavior of CO2 on a K-promoted HTlc. This model combines adsorption, diffusion, and reaction together with a Langmuir-Hinshelwood approach to describe the uptake and release processes of CO2 on a K-promoted HTlc. Results are presented for the first time that

10.1021/ie061042k CCC: $37.00 © 2007 American Chemical Society Published on Web 02/13/2007

1738

Ind. Eng. Chem. Res., Vol. 46, No. 6, 2007

convincingly reveal the adsorption and desorption behavior of CO2 on a K-promoted HTlc to be associated with complex, highly coupled, completely reversible adsorption, diffusion, and reaction phenomena. Adsorbent Material Preparation and Cycling An HTlc with molecular formula [Mg3Al(OH)8]2CO3‚nH2O was prepared by a coprecipitation method.1 While vigorously stirring, a solution of 41.7 mL of deionized water containing 0.75 mol Mg(NO3)2‚6H2O and 0.25 mol Al(NO3)3‚9H2O was added to a solution of 83.3 mL of deionized water containing 1.7 mol NaOH and 0.5 mol Na2CO3. The precipitate was separated from the slurry by vacuum filtration. The wet filter cake was washed with deionized water and was vacuum filtered three times, dried overnight at 60 °C in a vacuum oven, crushed, and calcined in air at 400 °C for 4 h. A K-promoted HTlc with molecular formula [Mg3Al(OH)8]2CO3‚K2CO3‚nH2O was prepared using an incipient wetness procedure. To obtain an Al:K ratio of 1:1, a 0.33 M solution of K2CO3 was prepared in deionized water, and a predetermined volume of it was added to the HTlc powder in three steps: (1) The solution was added dropwise to the powder until it appeared wet. (2) The wet powder was dried for 15 min in a vacuum oven at 60 °C. (3) Steps 1 and 2 were repeated until all the solution was added. A Perkin-Elmer TGA-7 thermogravimetric analyzer was used to measure the dynamic adsorption and desorption behavior of CO2 on this K-promoted HTlc. A typical TGA run was carried out with a sample of K-promoted HTlc powder (∼35 mg). This powder had a rather broad particle size distribution, with particle diameters ranging from 5 to 100 µm. First, the sample was activated at 400 °C for a specified length of time (8, 12, 16, or 20 h) in helium flowing at about 60 cm3/min and 1 atm. At the end of the activation step, the temperature was maintained at 400 °C and the gas was switched from helium to CO2 (also flowing at about 60 cm3/min and 1 atm) to initiate adsorption and begin the first half of an adsorption-desorption cycle. This adsorption half-step was continued for a specified length of time and then the gas was switched back to helium to initiate desorption and to finish the second half of the adsorption-desorption cycle. When the half-cycle time was set at a very long time of 700 min, one adsorption-desorption cycle was carried out to allow the system to approach equilibrium at the end of the adsorption or desorption step. When the halfcycle times were set at much shorter times of 15, 30, 45, 60, and 75 min, four adsorption-desorption cycles were carried out to elucidate the dynamic behavior during cycling under more reasonable cycle times that were far away from equilibrium. All of the CO2 loadings on K-promoted HTlc were based on the weight of the sample at the end of the activation period. However, at the end of this period, the sample was not necessarily in an equilibrium state; it was still losing weight at a slow, but noticeably steady, rate without showing any sign of leveling off, and it still contained some undesorbed CO2.16 Because this slow activation (desorption) rate would have required inordinately long activation periods to reach a more activatedsbut not necessarily a completely activatedsstate, the activation time was limited to durations between 8 and 20 h. Kinetic Model Development The nonequilibrium kinetic model was developed through a painstaking effort that involved starting with the simplest formulation and adding complexity until a model was found

Figure 1. Reaction pathway that describes the reversible adsorption and desorption behavior of CO2 in a K-promoted HTlc.

that satisfactorily predicted the long cycle time experiment. Hence, this modeling effort began with a simple isothermal solid diffusion model, that is, a linear driving force (LDF) approach, that has been used in the literature for this system.4-12 It failed to predict the long cycle time experiment, however. This isothermal model was extended to include both pore and surface diffusion and pore and loading dependent surface diffusion phenomena with no avail. Film mass transfer outside the K-promoted HTlc particles was also added to all of these models. As expected, its effects were short-lived and also did not play a role. These models were all evaluated again after a non-isothermal energy balance was included in the formulation with similar disappointing results. An isothermal solid diffusion model with one-solid-phase reaction also did not satisfactorily fit the long cycle time experiment. Finally, through careful examination of the behavior of the K-promoted HTlc when characterized in terms of nonequilibrium dynamic isotherms,16 and by taking into account the increasing level of complexity needed to fit the long cycle time experiment, it became apparent that one fast and two slow processes were taking place. Hence, an isothermal model with three reactions was devised. Two of the reactions were of the Langmuir-Hinshelwood type and one of them was a mass-transfer-limited (LDF) chemisorption process. This formulation is explained below. On the basis of recent findings in the literature16 and the painstaking modeling effort just described, the reaction pathway depicted in Figure 1 was envisioned for the reversible uptake and release of CO2 in K-promoted HTlc. This reaction pathway involves three reversible reactions with slow, intermediate, and fast adsorption and desorption behavior, respectively. It also involves four phases that participate in these reactions, each one represented by a reaction site and denoted by letters A, B, C, and E. A possible chemical formulation for each of these phases is given in Figure 1. An irreversible step, consisting of dehydration and dehydroxilation, which transforms the inactive sites I into active sites C, is also presented in the figure. This irreversible step represents the activation step of the K-promoted HTlc sample, which occurs only once. Reaction sites corresponding to phases C, B, and E are assumed to be in the form of carbonates. Phases C, B, and E, respectively, represent sites where two, one, and zero molecules of CO2 can strongly and chemically bind to the structure of the K-promoted HTlc. These three sites combined also represent the total number of strong reaction sites available and are denoted by qT. In contrast, phase A represents a site that weakly binds CO2 through a chemisorption mechanism. Phase A also actively participates in the formation of both phases B and C and, through a mass-transfer-limited (i.e., diffusional) process, in the rapid conversion between gaseous and chemisorbed CO2.

Ind. Eng. Chem. Res., Vol. 46, No. 6, 2007 1739

The reaction pathway depicted in Figure 1 is defined such that all the reactions proceed from phases with higher to lower CO2 contents, that is, in the direction of CO2 desorption or release. The first two reversible reactions represent the conversion of phase C into phase B and then the conversion of phase B into phase E, in each case giving up a CO2 species to form phase A. The third reversible reaction represents the conversion of phase A into gaseous CO2, which diffuses through the K-promoted HTlc particle according to the mass-transfer-limited LDF process. The reversible, nonequilibrium, kinetic model consists of three overall differential equations that represent the mass balances for the three reversible phases C, B, and A, and one algebraic equation that represents the mass balance restriction involvingphases C, B, and E. These four relationships are written as follows:

dqC ) - k1,fqC + k1,bqAqB dt

(1)

dqB ) k1,fqC - k1,bqAqB - k2,fqB + k2,bqAqE dt

(2)

dqA ) km(qA,e - qA) + k1,fqC - k1,bqAqB + k2,fqB - k2,bqAqE dt (3) qE ≡ q T - q B - q C

(4)

qA, qB, qC, and qE, respectively, represent the site concentrations of phases A, B, C, and E, with CO2-free K-promoted HTlc (i.e., Mg6Al2K2O10) as the basis. k1,f and k1,b represent the forward and backward rate constants for the first reversible reaction in Figure 1. k2,f and k2,b represent the forward and backward rate constants for the second reversible reaction in Figure 1. km represents the mass-transfer coefficient for the process involving site A and gaseous CO2 becoming chemisorbed within the particle. In this mass-transfer process, the LDF is defined between qA and qA,e, where qA,e represents the value of qA at equilibrium with T and PCO2, which is an independent parameter. Depending on whether the sample is under adsorption (qA,e ) qA,e,a > qA) or desorption (qA,e ) qA,e,d < qA), the mass-transfer coefficient km of this process takes on one of the following values:

{

k q )q >q km ) km,a qA,e ) qA,e,a < qA m,d A,e A,e,d A

}

(5)

The adsorption and desorption mass-transfer coefficients were allowed to be different to account for the possibility of a loadingdependent mass-transfer process. Finally, the CO2 loading is readily defined in terms of qA, qB, and qC as

qCO2 ≡ (qA - qA,o) + (qB - qB,o) + 2(qC - qC,o)

(6)

where qAo, qBo, and qCo, respectively, represent the site concentrations of phases A, B, and C just after activation. The CO2 release and uptake model described by eqs 1-6 constitutes two kinetically slow Langmuir-Hinshelwood processes coupled with a fast mass-transfer chemisorption process. This model has 12 parameters, namely, km,a, km,d, k1,f, k1,b, k2,f, k2,b, qA,o, qB,o, qC,o, qT, qA,e,a, and qA,e,d. However, values for three of these parameters (i.e., qB,o, qA,o, and qA,e,d) are known implicitly from sound assumptions, and a value for one of them (i.e., qT) can be calculated. The remaining eight parameters are

fitting parameters. The methodology used to obtain values for all of these parameters is explained in detail below. Results and Discussion Interpretation of Experimental Data. All the experimental CO2 loadings reported in Figures 2-4 are in terms of the CO2free K-promoted HTlc basis (i.e., Mg6Al2K2O10), according to

qCO2,exp )

x - xo (1 + qA,o + qB,o + 2qC,o) xo

(7)

where x represents the experimental mass obtained from the TGA and xo represents the value of x just after activation. Figure 2 displays the behavior of CO2 in K-promoted HTlc at 400 °C during a single adsorption and desorption cycle with a 700min half-cycle time. Figure 2a displays the full 1400-min cycle. Figure 2b and 2c, respectively, displays only the first 20 min of the adsorption and desorption steps to exemplify the behavior at short times. Recall that the loadings displayed in Figure 2 were normalized relative to the weight of the sample at the end of the activation step, which as indicated earlier, was not necessarily at equilibrium and still contained some CO2 in the sample. After reaching a CO2 loading of qCO2 ) 1.62 mol/kg at the end of the 700-min adsorption step, the sample returned to its original state with a qCO2 ≈ 0.0 mol/kg after a 700-min desorption step (Figure 2a). This long cycle time experiment clearly demonstrated the complete reversibility of CO2 in K-promoted HTlc. Another interesting feature shown in Figure 2 was the steady rate (i.e., linear trend) exhibited by the CO2 loading during the later stages of both the adsorption and desorption steps. In fact, for the desorption step, the trend was so steady that the CO2 loading would have become negative if a longer time was allowed for desorption. A negative loading would correspond to the sample weighing less and being more activated than the reference state. Hence, this behavior was similar to that observed at the end of the activation step (results not shown), which was an important observation that indicated the sample was still far removed from equilibrium and that it was undergoing the same chemical transformation that was characterized by the same reversible but very slow kinetics in both cases. It was further surmised that the process responsible for this slow and steady behavior observed during the later stages of desorption was also responsible for the slow and steady behavior observed during the later stages of adsorption, which defined the reversible nature of this particularly slow kinetic process. This very slow kinetic behavior contrasted significantly with the very fast kinetic behavior observed during the early stages of both the adsorption and desorption steps, as highlighted in Figure 2b and 2c. This very fast process at short times indicated that the sample was undergoing an altogether different chemical transformation compared to the very slow process at long times. The large difference between the rates of adsorption and desorption was also very apparent during these early stages, with the former being about 10 times faster than the latter. These rate differences were in good agreement with those reported by Ding and Alpay6 and signified that the sample was perhaps undergoing a loading-dependent mass-transfer process, where the CO2 was only weakly adsorbed on the K-promoted HTlc. However, neither the very fast nor the very slow kinetic processes could explain the adsorption and desorption behavior of CO2 in K-promoted HTlc during the intermediate stages that ensued between 5 and 300 min of both the adsorption and desorption steps. It was surmised that the adsorption and

1740

Ind. Eng. Chem. Res., Vol. 46, No. 6, 2007

Figure 2. TGA experimental run (empty circles) and model fit (thick solid line labeled qCO2) of the CO2 loading in a K-promoted HTlc and model predictions of the site concentrations of phases A, B, and C (thin solid lines) during one CO2 adsorption and desorption cycle with a 700-min halfcycle time at 400 °C: (a) complete 1400-min adsorption and desorption cycle, (b) first 20 min of the adsorption step, and (c) first 20 min of the desorption step. This sample of K-promoted HTlc was activated in helium at 400 °C for 12 h.

desorption behavior of the sample during those stages was controlled by a third and altogether different mechanism than the previous two mechanisms. It was further assumed that this mechanism was probably similar in nature to the very slow kinetic process but with somewhat faster kinetics. This intermediate behavior was consistent with the results reported recently by Ebner et al.16 It was becoming very clear that all three mechanisms collectively supported the existence of the four different phases identified in Figure 1 as phases A, B, C, and E. First, the slow process observed during the later stages of both the adsorption

and desorption steps, as well as during the last moments of the activation step (not shown), described the slow and reversible decomposition of phase C, which contained two CO2 molecules, into an intermediate phase B, which contained only one CO2 molecule, and phase A, which possessed a weakly bound CO2 molecule. The CO2 molecules in phases B and C were most likely in the form of carbonates, as depicted. The very fast process, on the other hand, was presumably because of phase A, which consisted of chemisorbed CO2 molecules diffusing through the structure of the K-promoted HTlc and reversibly converting into gaseous CO2 via a fast mass-transfer-limited process. Finally, the intermediate process observed during the intermediate stages of the adsorption and desorption steps (i.e., between 5 and 300 min) was probably associated with the reversible decomposition of phase B into phase A and a phase devoid of CO2 molecules, that is, phase E. Model Calibration with Experimental Data. These three kinetic processes that vividly explained the uptake and release of CO2 in K-promoted HTlc were illustrated mathematically by the reversible, nonequilibrium, kinetic model given by eqs 1-6. The excellent fit of this model to this long cycle time adsorption-desorption run over the entire 1400-min duration of the experiment is also shown in Figure 2. The model was also used to predict the site concentrations of phases A, B, and C throughout the experiment. These results are also plotted in Figure 2. The 12 model parameters that provided this excellent fit of the long cycle time TGA run and the corresponding predictions of the site concentration profiles were obtained as follows. As stated above in the Kinetic Model Development section, four of the parameters were known a priori, which left only eight of them to be determined by fitting the model to experimental data. The first two known parameters were qB,o and qA,o. It was easy to reason that at the end of the activation step only site C still contained some CO2 molecules as a result of the extremely slow kinetics involved with the conversion of this phase into phases B and A. In contrast, the much faster kinetics involved with the conversion of phase B into phase E and especially the conversion of phase A into gaseous CO2 granted that qB,o ) qA,o ) 0 for all time (t g 0) after activation. The third known parameter was qA,e,d. For similar reasons as given for qB,o and qA,o, it was easy to reason that the equilibrium condition for site A at the end of a desorption step was a vacant site free of CO2 molecules. It then followed that qA,e,d ) 0 for all time (t g 0) after activation. The fourth known parameter was qT. This parameter was calculated from stoichiometry by first considering a fully activated form of K-promoted HTlc that contained no CO2 molecules. This CO2 free state corresponded to phase E with molecular formula Mg6Al2K2O10. The maximum number of CO2 molecules that could be held by this K-promoted HTlc corresponded to phase C being completely filled with them, which was achieved by adding CO2 molecules to phase E until phase C was completely saturated. This state corresponded to molecular formula Mg6Al2K2O10(CO3)2 and resulted in qT ) 2.283 mol/kg. With four of the parameters known, the single-cycle adsorption-desorption TGA run depicted in Figure 2 with a 700-min half-cycle time was used to determine the remaining eight parameters in the model. This was accomplished in a sequential manner because, of the eight parameters, three were associated with the formation of phase A (i.e., km,a, km,d, and qA,e,a), two were associated with the formation of phase B (i.e., k1,f, and k1,b), and three were associated with the formation of phase C

Ind. Eng. Chem. Res., Vol. 46, No. 6, 2007 1741 Table 1. Assumed, Calculated, and Fitted Parameters Used in the Reversible, Nonequilibrium, Kinetic Model That Describe the Adsorption and Desorption Behavior of CO2 in a K-Promoted HTlc at 400 °C parameter

value

km,a km,d k1,f k1,b k2,f k2,b

1.218 × 100 min-1 1.397 × 10-1 min-1 1.600 × 10-4 min-1 1.222 × 10-3 kg mol-1 min-1 2.192 × 10-2 min-1 5.793 × 10-2 kg/mol-1 min-1

qA,e,a qA,e,d qT qA,o qB,o qC,o

0.932 mol/kg 0.000 mol/kg 2.283 mol/kg 0.000 mol/kg 0.000 mol/kg 1.587 mol/kg

(i.e., k2,f, k2,b, and qC,o). In this way, these three sets of parameters were uniquely independent in that they defined the behavior of their respective reaction with little influence over the other reactions. The values of all 12 parameters are listed in Table 1. The results in Figure 2 showed that there was very good agreement between the model and the experiments. For the adsorption step, only minor deviations between the model and the experiment resulted with the largest, but still acceptable, difference occurring between times of 3 and 20 min, as shown in Figure 2b. The same was true for the desorption step, with the largest difference occurring at times greater than about 1300 min, as shown in Figure 2a. Between times of 1300 and 1400 min, the model underpredicted the experimental CO2 loading and actually predicted slightly negative values of no more than -0.03 mol/kg. As explained earlier, these perfectly feasible negative values from the model indicated that the sample was losing more CO2 than that corresponding to its loading just after activation. The very good agreement between the model and the experiments lent some credence to the magnitudes of the parameters obtained from the model. For example, from the values of the adsorption (1.218 min-1) and desorption (0.140 min-1) mass-transfer coefficients, the model showed that the formation of phase A from gaseous CO2 was a relatively fast process and that the reverse of this process was somewhat slower. This order of magnitude difference in the adsorption and desorption mass-transfer coefficients was in agreement with that reported by Ding and Alpay;6 however, their magnitudes were both larger and more similar to those reported by Soares et al.8 The resulting values of the rate constants for the slowest reaction, that is, k1,f and k1,b, with respective values of 1.600 × 10-4 min-1 and 1.222 × 10-3 kg min-1 mol-1, corroborated that the dynamics of the reversible conversion of phase C into phases B and A was an extremely slow process that developed over a period of hundreds of hours. This result indicated that the sample would reach a true equilibrium state only after an exorbitantly long time. The model also predicted that phase B, with intermediate kinetic rate parameters of 2.192 × 10-2 min-1 and 5.793 × 10-2 kg min-1 mol-1, respectively, for k2,f and k2,b, barely existed after the first 400 min of desorption. This result verified the assumption made about qB,o ) 0 at the end of the activation period. The model predictions of the site concentrations of phases A, B, and C shown in Figure 2 displayed trends that were consistent with three coupled reactions that have drastically different rate constants. For example, the model showed that during 700 min of adsorption and 700 min of desorption, the CO2 loading in phase C changed only marginally because of its inherently slow kinetics. It exhibited only slight increases during adsorption and similarly slight decreases during desorp-

tion. In contrast, the model showed that phase A quickly saturated during the first 5 min of the adsorption step and nearly disappeared during the first 20 min of the desorption step. This trend was distinctly characteristic of the rapid adsorption and slightly slower desorption phenomena associated with this weakly chemisorbed phase of CO2 in the K-promoted HTlc structure that was mass-transfer-limited in its conversion to gaseous CO2. The intermediate rate process associated with the creation and extinction of phase B not surprisingly exhibited a maximum during the adsorption step, as it was being rapidly created from the conversion of phase A into phase B and slowly depleted from the subsequent conversion of phase B into phase C. During desorption, phase B essentially disappeared, but only long after phase A disappeared, because of the significant difference in their kinetic rate processes. Model Validation with Experimental Data. Without any further adjustments, these 12 parameters were then used to further validate and evaluate the model by predicating the behavior of the four-cycle adsorption-desorption TGA runs with half-cycle times of 15, 30, 45, 60, and 75 min. The experimental and modeling results are shown in Figure 3 for each cycle time, along with the predictions of the site concentrations of phases A, B, and C. All of the curves were characterized by a fast initial increase in the CO2 loading that was consistent with the fast initial rate of adsorption in the long cycle time run shown in Figure 2. Because of the slower initial rate of desorption (compared to adsorption) that was associated with phase A, none of the samples returned to the reference state defined at the end of the activation step, that is, qCO2 ) 0 mol/ kg. They did come increasingly closer to this state as the cycle time increased, however. With the longer cycle times increasingly giving more time for phase B to get involved in the CO2 uptake and release processes, this result was understood to indicate that the forward and reverse reaction rates associated with phase B were similar in magnitude, compared to the order of magnitude slower desorption rate than adsorption rate associated with phase A. The adsorption and desorption curves in Figure 3 all exhibited similar behaviors, not only from cycle to cycle and for the different cycle times, but also when compared to the long cycle time 1400-min curves shown in Figure 2. In particular, all the samples displayed the same behavior during the later stages of the adsorption and desorption steps, where the rate in all cases tended to approach a steady (i.e., linear) behavior, just like the curves in Figure 2. This result showed that, in all cases, an equilibrium state was once again far from being achieved. Nevertheless, the reversible nature of the adsorption and desorption behavior of CO2 in K-promoted HTlc was quite apparent even from these short cycle time results. Indeed, if a longer time was provided for desorption, each sample would have returned to its reference state. The results in Figure 3 also showed that the samples came very close to achieving periodic behavior just after the first cycle step (except for the sample cycled with a half-cycle time of 15 min, which is explained later). Almost constant CO2 working capacities were obtained from cycle to cycle that increased with increasing cycle time. Values of the CO2 working capacities ranged between 0.58 and 0.62, between 0.73 and 0.77, between 0.85 and 0.87, between 0.91 and 0.94, and between 0.99 and 1.05 mol/kg for respective half-cycle times of 15, 30, 45, 60, and 65 min. These subtle experimental behaviors observed with cycling were also simulated fairly accurately with the model. In all cases, the model provided excellent predictions of the first adsorption step. However, it exhibited some discrepancies

1742

Ind. Eng. Chem. Res., Vol. 46, No. 6, 2007

Figure 3. TGA experimental runs (empty circles) and model predictions (thick solid lines labeled qCO2) of the CO2 loadings in a K-promoted HTlc and model predictions of the site concentrations of phases A, B, and C (thin solid lines) during four CO2 adsorption and desorption cycles with (a) 15-, (b) 30-, (c) 45-, (d) 60-, and (e) 75-min half-cycle times at 400 °C. This sample of K-promoted HTlc was activated in helium at 400 °C for 12 h.

for all subsequent adsorption and desorption steps, but only during the later stages of those steps. For example, the model overpredicted the experimental working capacities by overpredicting the CO2 loadings during the adsorption steps and

underpredicting the CO2 loadings during the desorption steps; however, it did predict the initial rates of adsorption and desorption in all cases, even after four cycles. These discrepancies were perhaps because of a somewhat large equilibrium

Ind. Eng. Chem. Res., Vol. 46, No. 6, 2007 1743

loading of qA,e,a ) 0.932 mol/kg for the adsorption step, which might have been a consequence of the oversimplified approach used to predict the mass-transfer process between phase A and gaseous CO2. Using the site-concentration profiles of phases A, B, and C, the model helped explain the significant difference between the initial adsorption step and those of subsequent cycles. Clearly, phase C did not contribute to this behavior, as its CO2 loading did not change significantly during any of the runs, increasing only slightly after each cycle and oscillating more noticeably with longer cycle times (Figure 3d and 3e). This different behavior between the first and subsequent cycles was mainly due to phase B, with only a slight contribution from phase A. Both phases A and B displayed the largest increases in their CO2 loadings during the first adsorption step. The initial CO2 loadings observed experimentally were a result of these increases. The difference observed during the first adsorption step and those in subsequent ones tended to disappear for phase A with longer cycle times (Figures 3d and 3e). In contrast, the CO2 loading of phase B became increasingly less periodic as the cycle time decreased. In fact, for the shortest half-cycle time experiment of 15 min (Figure 3a), it continued to increase at the end of the adsorption step with cycling. This dynamic behavior associated with the slow kinetics of phase B during cycling at short times (Figure 3a) was not only predicted by the model, but it was also observed experimentally. The ability of phase B and, to a lesser extent, phase A, to exhibit higher initial uptakes of CO2 only during the first adsorption step also indicated that the working capacity of the K-promoted HTlc was independent of the activation period. This interesting result was studied both experimentally and theoretically by using the model to predict the two-cycle adsorptiondesorption TGA runs with 45-min half-cycle times and activation periods of 8, 12, 16, and 20 h. For this case, the values of qC,o necessarily changed according to the number of sites of phase C remaining after activation. These new values of qc,o were obtained simply by knowing the desorption rate k1,f of CO2 from phase C that was predicted from the model using the following expression:

qC,o ) qC,o,12h exp[-60k1,f(tact - 12)]

(8)

where tact represents the activation time in h and qC,o,12h represents the value of qC,o for the 12-h activation period case, which is given in Table 1. The resulting values of qC,o were 1.649, 1.587, 1.527, 1.470 mol/kg for the 8-, 12-, 16-, and 20-h activation periods, respectively. Figure 4 compares these experimental results to the model predictions. It also shows the predictions of the site concentrations of phases A and B for each activation period. In all cases, the reference loading was set to zero at the end of the activation period. The results in Figure 4 showed that both the model and the experiments always reached periodic behavior with similar working capacities, regardless of the activation period of the four samples. Although there were some obvious, but nevertheless acceptable, discrepancies between the model predictions and the experimental results, especially during the desorption step, the model predicted quite well the experimental trends associated with different activation times. It was interesting that the model predicted identical site concentration profiles for phase A no matter the activation period, which was indicative of the very fast reversible kinetic process associated with phase A. Phase B, on the other hand, exhibited increasing maxima of about 0.44, 0.48, 0.51, and 0.55 mol/kg with increasing activation times of 8, 12, 16, and 20 h, respectively. The working

Figure 4. TGA experimental runs (empty symbols) after 8 h (triangles), 12 h (squares), 16 h (diamonds), and 20 h (circles) of activation in helium at 400 °C and model predictions (thick solid line labeled qCO2) of the CO2 loadings in K-promoted HTlc and model predictions of the site concentrations of phases A and B (thin solid lines) during two CO2 adsorption and desorption cycles with a 45-min half-cycle time.

capacity of phase B also increased with the activation time, although not significantly. For example, it increased from 0.21 to 0.25 mol/kg when the activation time increased from 8 to 20 h. As expected, during cycling, the profiles for phase C (not shown) did not change appreciably from their respective qC,o values associated with each activation period; in fact, they looked like those shown in Figure 3c. Conclusions A nonequilibrium kinetic model was developed to describe the reversible adsorption and desorption behavior of CO2 in a K-promoted hydrotalcite-like compound (HTlc) at 400 °C. The model consisted of three reversible reactions and four phases. Two of the reactions were of the Langmuir-Hinshelwood type with slow and intermediate kinetics, and one was a masstransfer-limited chemisorption process with very fast kinetics. The first two reversible reactions represented the conversion of phase C (Mg6Al2K2O10(CO3)2) into phase B (Mg6Al2K2O9(CO3)) and then the conversion of phase B into phase E (Mg6Al2K2O10). In each case, a CO2 molecule was given up to form phase A (chemisorbed CO2). The third reversible reaction represented the conversion of phase A into gaseous CO2, which diffused through the K-promoted HTlc particle according to a mass-transfer-limited, linear driving force process. This model contained 12 parameters, four of which were known a priori. To calibrate and test this model, a K-promoted HTlc was synthesized according to a recipe in the literature and then was studied to determine its dynamic behavior during CO2 adsorption and desorption cycles carried out at 400 °C. Three sets of dynamic cycling experiments were carried out. A long cycle time adsorption (700 min) and desorption (700 min) experiment was carried out that approached equilibrium with CO2 in K-promoted HTlc at the end of each step. Prior to cycling in CO2 and helium, this sample was activated at 400 °C for 12 h in helium. Then, the effect of the half-cycle time (15, 30, 45, 60, and 75 min) was studied with samples activated for 12 h in helium at 400 °C and cycled four times each. Finally, the effect of the activation time (8, 12, 16, and 20 h) was studied with samples cycled twice with a 45-min half-cycle time.

1744

Ind. Eng. Chem. Res., Vol. 46, No. 6, 2007

The eight-parameter nonequilibrium kinetic model was fitted successfully to the long cycle time adsorption (700 min) and desorption (700 min) experiment. The model then predicted successfully the dynamic and cyclic behavior of both the much shorter cycle time experiments and the different activation time experiments. In all cases, it very accurately simulated the reversible adsorption and desorption behavior of the very fast mass-transfer-limited process associated with phase A, and the intermediate and slow kinetic processes associated with phases B and C, respectively. It also approached periodic behavior during cycling for different half-cycle times in the same way the experiments did; also, it showed independence between the CO2 working capacity and the activation time of the sample, which was also consistent with the experimental findings. Overall, this nonequilibrium kinetic model revealed for the first time that the adsorption and desorption behavior of CO2 in a K-promoted HTlc was associated with a combination of completely reversible adsorption, diffusion, and reaction phenomena. Acknowledgment The authors gratefully acknowledge financial support provided by DOE through Grant No. DE-FG26-03NT41799 and MeadWestvaco and the Separations Research Program at the University of Texas at Austin. Nomenclature A ) reaction site for weakly chemisorbed CO2 in a K-promoted HTlc B ) reaction site for one molecule of chemically bound CO2 in a K-promoted HTlc, i.e., Mg6Al2K2O9(CO3) C ) reaction site for two molecules of chemically bound CO2 in a K-promoted HTlc, i.e., Mg6Al2K2O8(CO3)2 E ) reaction site free of CO2 in a K-promoted HTlc, i.e., Mg6Al2K2O10 I ) reaction site for two molecules of chemically bound CO2 prior to dehydration and dehydroxylation, i.e., [Mg3Al(OH)8](CO3)‚K2CO3‚nH2O k1,f ) forward rate constant for the reaction C T A + B (min-1) k1,b ) backward rate constant for the reaction C T A + B (kg/ mol/min) k2,f ) forward rate constant for the reaction B T A + E (min-1) k2,b ) backward rate constant for the reaction B T A + E (kg/ mol/min) km,a ) mass-transfer coefficient for adsorption of CO2 from the gas phase into phase A (min-1) km,d ) mass-transfer coefficient for desorption of CO2 from phase A into the gas phase (min-1) qA,e ) concentration of site A at equilibrium (mol of sites/kg of CO2-free K-promoted HTlc) qA,e,a ) concentration of site A at equilibrium during adsorption (mol of sites/kg of CO2-free K-promoted HTlc) qA,e,d ) concentration of site A at equilibrium during desorption (mol of sites/kg of CO2-free K-promoted HTlc) qCO2 ) CO2 loading relative to qo (mol CO2/kg of CO2-free K-promoted HTlc) qo ) CO2 loading after activation (mol CO2/kg of CO2-free K-promoted HTlc)

qT ) total number of reaction sites available for chemically bound CO2 (mol of sites/kg of CO2-free K-promoted HTlc) qX ) concentration of site X, with X ) A, B, C, or E (mol of sites/kg of CO2-free K-promoted HTlc) qX,o ) concentration of site X after activation, with X ) A, B, C, or E (mol of sites/kg of CO2-free K-promoted HTlc) tact ) activation time (h) x ) experimental mass obtained from the TGA (g) xo ) experimental mass obtained from the TGA just after activation (g) Literature Cited (1) Nataraj, S.; Carvill, B. T.; Hufton, J. R.; Mayorga, S. G.; Gaffney, T. R.; Brzozowski, J. R. Process for Operating Equilibrium Controlled Reactions. Can. Patent 2,235,928, 1998. (2) Hufton, J. R.; Mayorga, S.; Sircar, S. Sorption-Enhanced Reaction Process for Hydrogen Production. AIChE J. 1999, 45, 248-256. (3) Hufton, J. R.; Allam, R. J.; Chiang, R.; Middleton, P.; Weist, E. L.; White V. Development of a Process for CO2 Capture Gas Turbines using a Sorption Enhamced Water Gas Shift Reactor System. Presented at the 7th International Conference on Green House Control Technologies, Vancouver, Canada, 2004. (4) Ding, Y.; Alpay, E. Equilibria and Kinetics of CO2 Adsorption on Hydrotalcite Adsorbent. Chem. Eng. Sci. 2000, 55, 3461-3474. (5) Ding, Y.; Alpay, E. Adsorption-Enhanced Steam-Methane Reforming. Chem. Eng. Sci. 2000, 55, 3929-3940. (6) Ding, Y.; Alpay, E. High Temperature Recovery of CO2 from Flue Gases Using Hydrotalcite Adsorbent. Trans. Inst. Chem. Eng. 2001, 79, 45-51. (7) Yong, Z.; Mata, V.; Rodrigues, A. E. Adsorption of Carbon Dioxide onto Hydrotalcite-Like Compounds (HTlcs) at High Temperature. Ind. Eng. Chem. Res. 2001, 40, 204-209. (8) Soares, J. L.; Grande, C. A.; Yong, Z; Moreira, R. F. P. M.; Rodrigues, A. E. Adsorption of Carbon Dioxide at High Temperatures onto Hydrotalcite-Like Compounds (HTlcs). In Fundamentals of Adsorption 7; Kaneko, K., Kanoh, H., Hanzawa, Y., Eds.; IK International, Ltd.: ChibuCity, Japan, 2002; pp 763-770. (9) Soares, J. L.; Moreira, R. F. P. M.; Jose, H. J.; Grande, C. A.; Rodrigues, A. E. Hydrotalcite Materials for Carbon Dioxide Adsorption at High Temperatures: Characterization and Diffusivity Measurements. Sep. Sci. Technol. 2004, 39, 1989-2010. (10) Moreira, R. F. P. M.; Soares, J. L.; Casarin, G. L.; Rodrigues, A. E. Adsorption of CO2 on Hydrotalcite-Like Compounds in a Fixed Bed. Sep. Sci. Technol. 2006, 41, 341-357. (11) Reynolds, S. P.; Ebner, A. D.; Ritter, J. A. New Pressure Swing Adsorption Cycles for Carbon Dioxide Sequestration. Adsorption 2005, 11, 531-536. (12) Reynolds, S. P.; Ebner, A. D.; Ritter, J. A. Stripping PSA Cycles for CO2 Recovery from Flue Gas at High Temperature Using a HydrotalciteLike Adsorbent. Ind. Eng. Chem. Res. 2006, 4278-4294. (13) Yong, Z.; Mata, V.; Rodrigues, A. E. Adsorption of Carbon Dioxide at High Temperature-A Review. Sep. Purif. Technol. 2002, 26, 195-205. (14) Yong, Z.; Rodrigues, A. E. Hydrotalcite-Like Compounds as Adsorbents for Carbon Dioxide. Energy ConVers. Manage. 2002, 43, 18651876. (15) Hutson, N. D.; Speakman, S. A.; Payznat, E. A. Structural Effects on the High Temperature Adsorption of CO2 on a Synthetic Hydrotalcite. Chem. Mater. 2004, 16, 4135-4143. (16) Ebner, A. D.; Reynolds, S. P.; Ritter, J. A. Understanding the Adsorption and Desorption Behavior of CO2 on a K-Promoted HTlc through Non-Equilibrium Dynamic Isotherms. Ind. Eng. Chem. Res. 2006, 45, 63876392.

ReceiVed for reView August 8, 2006 ReVised manuscript receiVed December 11, 2006 Accepted January 11, 2007 IE061042K